>-----Original Message----- >From: Jan Kiszka <jan.kis...@siemens.com> >Sent: Wednesday, August 4, 2021 2:48 PM >To: Chen, Hongzhan <hongzhan.c...@intel.com>; xenomai@xenomai.org >Subject: Re: xenomai warning after upgrade to linux 5.4.115 > >On 04.08.21 02:32, Chen, Hongzhan wrote: >> >> >>> -----Original Message----- >>> From: Jan Kiszka <jan.kis...@siemens.com> >>> Sent: Tuesday, August 3, 2021 5:37 PM >>> To: Chen, Hongzhan <hongzhan.c...@intel.com>; xenomai@xenomai.org >>> Subject: Re: xenomai warning after upgrade to linux 5.4.115 >>> >>> On 29.07.21 10:02, Jan Kiszka via Xenomai wrote: >>>> On 29.07.21 08:43, Chen, Hongzhan wrote: >>>>> After we upgraded our linux to 5.4.115 from 5.4.76 and rebased IPIPE >>>>> patchset for xenomai 3.1.X, there is xenomai warning >>>>> printed as in attached xenomai_warning_on_linux5.4.115 on some specific >>>>> boards according to our validation. >>>> >>>> From a first inspection: Linux considers its interrupts to be off around >>>> >>>> xenomai_init >>>> xntimer_grab_hardware >>>> ipipe_timer_start >>>> smp_call_function_single >>>> >>>> But that should not be the case. The call path is valid and used for a >>>> long time. The issue should come from somewhere else. >>>> >>>> You you try to debug where interrupts are turned off from Linux >>>> perspective? Compare that to boot ups where Linux does not complain. >>>> >>>>> But I cannot reproduce the issue on same type of board like Maxtang that >>>>> I have. In addition, >>>>> on one specific board, we found that there is similar warning happened in >>>>> smp_call_function_single when ACPI init >>>>> ahead of warning happened in xenomai init as in attached >>>>> ACPI_warning_and_xenomai_warning_on_linux5.4.115. >>>>> When we set acpi=off in cmdline for all problematic boards , warning >>>>> totally disappear. Since I cannot get the problematic board >>>>> to further debug, would you mind help check attached logs and kernel >>>>> config and find the root cause out? >>>>> Thanks for your help in advance. >>>> >>>> >>>> Does that other case only occur when CONFIG_IPIPE is enabled? If it also >>>> happens with plain Linux, we may see a mainline issue. Or firmware bugs. >>>> >>>> Jan >>>> >>> >>> Any news on this issue? >> >> Finally , we found there is missing code in our split patches. >> > >Means, your kernel was missing something that the official I-pipe >patches have?
Yes. Our rebased IPIPE patchset miss one line of code for x2apic_cluster.c. > >Jan > >-- >Siemens AG, T RDA IOT >Corporate Competence Center Embedded Linux